#include "genxml/gen_macros.h"
#include "genxml/genX_pack.h"
-static uint32_t
-cmd_buffer_flush_push_constants(struct anv_cmd_buffer *cmd_buffer)
-{
- static const uint32_t push_constant_opcodes[] = {
- [MESA_SHADER_VERTEX] = 21,
- [MESA_SHADER_TESS_CTRL] = 25, /* HS */
- [MESA_SHADER_TESS_EVAL] = 26, /* DS */
- [MESA_SHADER_GEOMETRY] = 22,
- [MESA_SHADER_FRAGMENT] = 23,
- [MESA_SHADER_COMPUTE] = 0,
- };
-
- VkShaderStageFlags flushed = 0;
-
- anv_foreach_stage(stage, cmd_buffer->state.push_constants_dirty) {
- if (stage == MESA_SHADER_COMPUTE)
- continue;
-
- struct anv_state state = anv_cmd_buffer_push_constants(cmd_buffer, stage);
-
- if (state.offset == 0) {
- anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_CONSTANT_VS),
- ._3DCommandSubOpcode = push_constant_opcodes[stage]);
- } else {
- anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_CONSTANT_VS),
- ._3DCommandSubOpcode = push_constant_opcodes[stage],
- .ConstantBody = {
- .PointerToConstantBuffer2 = { &cmd_buffer->device->dynamic_state_block_pool.bo, state.offset },
- .ConstantBuffer2ReadLength = DIV_ROUND_UP(state.alloc_size, 32),
- });
- }
-
- flushed |= mesa_to_vk_shader_stage(stage);
- }
-
- cmd_buffer->state.push_constants_dirty &= ~flushed;
-
- return flushed;
-}
-
#if GEN_GEN == 8
-static void
-emit_viewport_state(struct anv_cmd_buffer *cmd_buffer,
- uint32_t count, const VkViewport *viewports)
+void
+gen8_cmd_buffer_emit_viewport(struct anv_cmd_buffer *cmd_buffer)
{
+ uint32_t count = cmd_buffer->state.dynamic.viewport.count;
+ const VkViewport *viewports = cmd_buffer->state.dynamic.viewport.viewports;
struct anv_state sf_clip_state =
anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, count * 64, 64);
- struct anv_state cc_state =
- anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, count * 8, 32);
for (uint32_t i = 0; i < count; i++) {
const VkViewport *vp = &viewports[i];
.YMaxClipGuardband = 1.0f,
.XMinViewPort = vp->x,
.XMaxViewPort = vp->x + vp->width - 1,
- .YMinViewPort = vp->y,
- .YMaxViewPort = vp->y + vp->height - 1,
- };
-
- struct GENX(CC_VIEWPORT) cc_viewport = {
- .MinimumDepth = vp->minDepth,
- .MaximumDepth = vp->maxDepth
+ .YMinViewPort = MIN2(vp->y, vp->y + vp->height),
+ .YMaxViewPort = MAX2(vp->y, vp->y + vp->height) - 1,
};
GENX(SF_CLIP_VIEWPORT_pack)(NULL, sf_clip_state.map + i * 64,
&sf_clip_viewport);
- GENX(CC_VIEWPORT_pack)(NULL, cc_state.map + i * 8, &cc_viewport);
}
- if (!cmd_buffer->device->info.has_llc) {
- anv_state_clflush(sf_clip_state);
- anv_state_clflush(cc_state);
- }
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_flush(sf_clip_state);
anv_batch_emit(&cmd_buffer->batch,
- GENX(3DSTATE_VIEWPORT_STATE_POINTERS_CC),
- .CCViewportPointer = cc_state.offset);
- anv_batch_emit(&cmd_buffer->batch,
- GENX(3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP),
- .SFClipViewportPointer = sf_clip_state.offset);
+ GENX(3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP), clip) {
+ clip.SFClipViewportPointer = sf_clip_state.offset;
+ }
}
void
-gen8_cmd_buffer_emit_viewport(struct anv_cmd_buffer *cmd_buffer)
+gen8_cmd_buffer_emit_depth_viewport(struct anv_cmd_buffer *cmd_buffer,
+ bool depth_clamp_enable)
{
- if (cmd_buffer->state.dynamic.viewport.count > 0) {
- emit_viewport_state(cmd_buffer, cmd_buffer->state.dynamic.viewport.count,
- cmd_buffer->state.dynamic.viewport.viewports);
- } else {
- /* If viewport count is 0, this is taken to mean "use the default" */
- emit_viewport_state(cmd_buffer, 1,
- &(VkViewport) {
- .x = 0.0f,
- .y = 0.0f,
- .width = cmd_buffer->state.framebuffer->width,
- .height = cmd_buffer->state.framebuffer->height,
- .minDepth = 0.0f,
- .maxDepth = 1.0f,
- });
- }
-}
-#endif
+ uint32_t count = cmd_buffer->state.dynamic.viewport.count;
+ const VkViewport *viewports = cmd_buffer->state.dynamic.viewport.viewports;
+ struct anv_state cc_state =
+ anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, count * 8, 32);
-static void
-emit_lri(struct anv_batch *batch, uint32_t reg, uint32_t imm)
-{
- anv_batch_emit(batch, GENX(MI_LOAD_REGISTER_IMM),
- .RegisterOffset = reg,
- .DataDWord = imm);
-}
+ for (uint32_t i = 0; i < count; i++) {
+ const VkViewport *vp = &viewports[i];
-#define GEN8_L3CNTLREG 0x7034
+ struct GENX(CC_VIEWPORT) cc_viewport = {
+ .MinimumDepth = depth_clamp_enable ? vp->minDepth : 0.0f,
+ .MaximumDepth = depth_clamp_enable ? vp->maxDepth : 1.0f,
+ };
-static void
-config_l3(struct anv_cmd_buffer *cmd_buffer, bool enable_slm)
-{
- /* References for GL state:
- *
- * - commits e307cfa..228d5a3
- * - src/mesa/drivers/dri/i965/gen7_l3_state.c
- */
+ GENX(CC_VIEWPORT_pack)(NULL, cc_state.map + i * 8, &cc_viewport);
+ }
- uint32_t val = enable_slm ?
- /* All = 48 ways; URB = 16 ways; DC and RO = 0, SLM = 1 */
- 0x60000021 :
- /* All = 48 ways; URB = 48 ways; DC, RO and SLM = 0 */
- 0x60000060;
- bool changed = cmd_buffer->state.current_l3_config != val;
-
- if (changed) {
- /* According to the hardware docs, the L3 partitioning can only be changed
- * while the pipeline is completely drained and the caches are flushed,
- * which involves a first PIPE_CONTROL flush which stalls the pipeline and
- * initiates invalidation of the relevant caches...
- */
- anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
- .TextureCacheInvalidationEnable = true,
- .ConstantCacheInvalidationEnable = true,
- .InstructionCacheInvalidateEnable = true,
- .DCFlushEnable = true,
- .PostSyncOperation = NoWrite,
- .CommandStreamerStallEnable = true);
-
- /* ...followed by a second stalling flush which guarantees that
- * invalidation is complete when the L3 configuration registers are
- * modified.
- */
- anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
- .DCFlushEnable = true,
- .PostSyncOperation = NoWrite,
- .CommandStreamerStallEnable = true);
-
- emit_lri(&cmd_buffer->batch, GEN8_L3CNTLREG, val);
- cmd_buffer->state.current_l3_config = val;
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_flush(cc_state);
+
+ anv_batch_emit(&cmd_buffer->batch,
+ GENX(3DSTATE_VIEWPORT_STATE_POINTERS_CC), cc) {
+ cc.CCViewportPointer = cc_state.offset;
}
}
+#endif
static void
__emit_genx_sf_state(struct anv_cmd_buffer *cmd_buffer)
cmd_buffer->state.pipeline->gen8.sf);
}
-#include "genxml/gen9_pack.h"
-static void
-__emit_gen9_sf_state(struct anv_cmd_buffer *cmd_buffer)
+void
+gen9_emit_sf_state(struct anv_cmd_buffer *cmd_buffer);
+
+#if GEN_GEN == 9
+
+void
+gen9_emit_sf_state(struct anv_cmd_buffer *cmd_buffer)
{
- uint32_t sf_dw[GENX(3DSTATE_SF_length)];
- struct GEN9_3DSTATE_SF sf = {
- GEN9_3DSTATE_SF_header,
- .LineWidth = cmd_buffer->state.dynamic.line_width,
- };
- GEN9_3DSTATE_SF_pack(NULL, sf_dw, &sf);
- /* FIXME: gen9.fs */
- anv_batch_emit_merge(&cmd_buffer->batch, sf_dw,
- cmd_buffer->state.pipeline->gen8.sf);
+ __emit_genx_sf_state(cmd_buffer);
}
+#endif
+
+#if GEN_GEN == 8
+
static void
__emit_sf_state(struct anv_cmd_buffer *cmd_buffer)
{
if (cmd_buffer->device->info.is_cherryview)
- __emit_gen9_sf_state(cmd_buffer);
+ gen9_emit_sf_state(cmd_buffer);
else
__emit_genx_sf_state(cmd_buffer);
}
-void
-genX(cmd_buffer_flush_state)(struct anv_cmd_buffer *cmd_buffer)
-{
- struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
- uint32_t *p;
+#else
- uint32_t vb_emit = cmd_buffer->state.vb_dirty & pipeline->vb_used;
+static void
+__emit_sf_state(struct anv_cmd_buffer *cmd_buffer)
+{
+ __emit_genx_sf_state(cmd_buffer);
+}
- assert((pipeline->active_stages & VK_SHADER_STAGE_COMPUTE_BIT) == 0);
+#endif
- config_l3(cmd_buffer, false);
+void
+genX(cmd_buffer_enable_pma_fix)(struct anv_cmd_buffer *cmd_buffer, bool enable)
+{
+ if (cmd_buffer->state.pma_fix_enabled == enable)
+ return;
- genX(flush_pipeline_select_3d)(cmd_buffer);
+ cmd_buffer->state.pma_fix_enabled = enable;
- if (vb_emit) {
- const uint32_t num_buffers = __builtin_popcount(vb_emit);
- const uint32_t num_dwords = 1 + num_buffers * 4;
+ /* According to the Broadwell PIPE_CONTROL documentation, software should
+ * emit a PIPE_CONTROL with the CS Stall and Depth Cache Flush bits set
+ * prior to the LRI. If stencil buffer writes are enabled, then a Render
+ * Cache Flush is also necessary.
+ *
+ * The Skylake docs say to use a depth stall rather than a command
+ * streamer stall. However, the hardware seems to violently disagree.
+ * A full command streamer stall seems to be needed in both cases.
+ */
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
+ pc.DepthCacheFlushEnable = true;
+ pc.CommandStreamerStallEnable = true;
+ pc.RenderTargetCacheFlushEnable = true;
+ }
- p = anv_batch_emitn(&cmd_buffer->batch, num_dwords,
- GENX(3DSTATE_VERTEX_BUFFERS));
- uint32_t vb, i = 0;
- for_each_bit(vb, vb_emit) {
- struct anv_buffer *buffer = cmd_buffer->state.vertex_bindings[vb].buffer;
- uint32_t offset = cmd_buffer->state.vertex_bindings[vb].offset;
+#if GEN_GEN == 9
- struct GENX(VERTEX_BUFFER_STATE) state = {
- .VertexBufferIndex = vb,
- .MemoryObjectControlState = GENX(MOCS),
- .AddressModifyEnable = true,
- .BufferPitch = pipeline->binding_stride[vb],
- .BufferStartingAddress = { buffer->bo, buffer->offset + offset },
- .BufferSize = buffer->size - offset
- };
+ uint32_t cache_mode;
+ anv_pack_struct(&cache_mode, GENX(CACHE_MODE_0),
+ .STCPMAOptimizationEnable = enable,
+ .STCPMAOptimizationEnableMask = true);
+ anv_batch_emit(&cmd_buffer->batch, GENX(MI_LOAD_REGISTER_IMM), lri) {
+ lri.RegisterOffset = GENX(CACHE_MODE_0_num);
+ lri.DataDWord = cache_mode;
+ }
- GENX(VERTEX_BUFFER_STATE_pack)(&cmd_buffer->batch, &p[1 + i * 4], &state);
- i++;
- }
+#elif GEN_GEN == 8
+
+ uint32_t cache_mode;
+ anv_pack_struct(&cache_mode, GENX(CACHE_MODE_1),
+ .NPPMAFixEnable = enable,
+ .NPEarlyZFailsDisable = enable,
+ .NPPMAFixEnableMask = true,
+ .NPEarlyZFailsDisableMask = true);
+ anv_batch_emit(&cmd_buffer->batch, GENX(MI_LOAD_REGISTER_IMM), lri) {
+ lri.RegisterOffset = GENX(CACHE_MODE_1_num);
+ lri.DataDWord = cache_mode;
}
- cmd_buffer->state.vb_dirty &= ~vb_emit;
-
- if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_PIPELINE) {
- /* If somebody compiled a pipeline after starting a command buffer the
- * scratch bo may have grown since we started this cmd buffer (and
- * emitted STATE_BASE_ADDRESS). If we're binding that pipeline now,
- * reemit STATE_BASE_ADDRESS so that we use the bigger scratch bo. */
- if (cmd_buffer->state.scratch_size < pipeline->total_scratch)
- anv_cmd_buffer_emit_state_base_address(cmd_buffer);
-
- anv_batch_emit_batch(&cmd_buffer->batch, &pipeline->batch);
-
- /* From the BDW PRM for 3DSTATE_PUSH_CONSTANT_ALLOC_VS:
- *
- * "The 3DSTATE_CONSTANT_VS must be reprogrammed prior to
- * the next 3DPRIMITIVE command after programming the
- * 3DSTATE_PUSH_CONSTANT_ALLOC_VS"
- *
- * Since 3DSTATE_PUSH_CONSTANT_ALLOC_VS is programmed as part of
- * pipeline setup, we need to dirty push constants.
- */
- cmd_buffer->state.push_constants_dirty |= VK_SHADER_STAGE_ALL_GRAPHICS;
+#endif /* GEN_GEN == 8 */
+
+ /* After the LRI, a PIPE_CONTROL with both the Depth Stall and Depth Cache
+ * Flush bits is often necessary. We do it regardless because it's easier.
+ * The render cache flush is also necessary if stencil writes are enabled.
+ *
+ * Again, the Skylake docs give a different set of flushes but the BDW
+ * flushes seem to work just as well.
+ */
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
+ pc.DepthStallEnable = true;
+ pc.DepthCacheFlushEnable = true;
+ pc.RenderTargetCacheFlushEnable = true;
}
+}
+
+static inline bool
+want_depth_pma_fix(struct anv_cmd_buffer *cmd_buffer)
+{
+ assert(GEN_GEN == 8);
+
+ /* From the Broadwell PRM Vol. 2c CACHE_MODE_1::NP_PMA_FIX_ENABLE:
+ *
+ * SW must set this bit in order to enable this fix when following
+ * expression is TRUE.
+ *
+ * 3DSTATE_WM::ForceThreadDispatch != 1 &&
+ * !(3DSTATE_RASTER::ForceSampleCount != NUMRASTSAMPLES_0) &&
+ * (3DSTATE_DEPTH_BUFFER::SURFACE_TYPE != NULL) &&
+ * (3DSTATE_DEPTH_BUFFER::HIZ Enable) &&
+ * !(3DSTATE_WM::EDSC_Mode == EDSC_PREPS) &&
+ * (3DSTATE_PS_EXTRA::PixelShaderValid) &&
+ * !(3DSTATE_WM_HZ_OP::DepthBufferClear ||
+ * 3DSTATE_WM_HZ_OP::DepthBufferResolve ||
+ * 3DSTATE_WM_HZ_OP::Hierarchical Depth Buffer Resolve Enable ||
+ * 3DSTATE_WM_HZ_OP::StencilBufferClear) &&
+ * (3DSTATE_WM_DEPTH_STENCIL::DepthTestEnable) &&
+ * (((3DSTATE_PS_EXTRA::PixelShaderKillsPixels ||
+ * 3DSTATE_PS_EXTRA::oMask Present to RenderTarget ||
+ * 3DSTATE_PS_BLEND::AlphaToCoverageEnable ||
+ * 3DSTATE_PS_BLEND::AlphaTestEnable ||
+ * 3DSTATE_WM_CHROMAKEY::ChromaKeyKillEnable) &&
+ * 3DSTATE_WM::ForceKillPix != ForceOff &&
+ * ((3DSTATE_WM_DEPTH_STENCIL::DepthWriteEnable &&
+ * 3DSTATE_DEPTH_BUFFER::DEPTH_WRITE_ENABLE) ||
+ * (3DSTATE_WM_DEPTH_STENCIL::Stencil Buffer Write Enable &&
+ * 3DSTATE_DEPTH_BUFFER::STENCIL_WRITE_ENABLE &&
+ * 3DSTATE_STENCIL_BUFFER::STENCIL_BUFFER_ENABLE))) ||
+ * (3DSTATE_PS_EXTRA:: Pixel Shader Computed Depth mode != PSCDEPTH_OFF))
+ */
- /* We emit the binding tables and sampler tables first, then emit push
- * constants and then finally emit binding table and sampler table
- * pointers. It has to happen in this order, since emitting the binding
- * tables may change the push constants (in case of storage images). After
- * emitting push constants, on SKL+ we have to emit the corresponding
- * 3DSTATE_BINDING_TABLE_POINTER_* for the push constants to take effect.
+ /* These are always true:
+ * 3DSTATE_WM::ForceThreadDispatch != 1 &&
+ * !(3DSTATE_RASTER::ForceSampleCount != NUMRASTSAMPLES_0)
*/
- uint32_t dirty = 0;
- if (cmd_buffer->state.descriptors_dirty)
- dirty = gen7_cmd_buffer_flush_descriptor_sets(cmd_buffer);
- if (cmd_buffer->state.push_constants_dirty)
- dirty |= cmd_buffer_flush_push_constants(cmd_buffer);
+ /* We only enable the PMA fix if we know for certain that HiZ is enabled.
+ * If we don't know whether HiZ is enabled or not, we disable the PMA fix
+ * and there is no harm.
+ *
+ * (3DSTATE_DEPTH_BUFFER::SURFACE_TYPE != NULL) &&
+ * 3DSTATE_DEPTH_BUFFER::HIZ Enable
+ */
+ if (!cmd_buffer->state.hiz_enabled)
+ return false;
- if (dirty)
- gen7_cmd_buffer_emit_descriptor_pointers(cmd_buffer, dirty);
+ /* 3DSTATE_PS_EXTRA::PixelShaderValid */
+ struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
+ if (!anv_pipeline_has_stage(pipeline, MESA_SHADER_FRAGMENT))
+ return false;
+
+ /* !(3DSTATE_WM::EDSC_Mode == EDSC_PREPS) */
+ const struct brw_wm_prog_data *wm_prog_data = get_wm_prog_data(pipeline);
+ if (wm_prog_data->early_fragment_tests)
+ return false;
+
+ /* We never use anv_pipeline for HiZ ops so this is trivially true:
+ * !(3DSTATE_WM_HZ_OP::DepthBufferClear ||
+ * 3DSTATE_WM_HZ_OP::DepthBufferResolve ||
+ * 3DSTATE_WM_HZ_OP::Hierarchical Depth Buffer Resolve Enable ||
+ * 3DSTATE_WM_HZ_OP::StencilBufferClear)
+ */
- if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_DYNAMIC_VIEWPORT)
- gen8_cmd_buffer_emit_viewport(cmd_buffer);
+ /* 3DSTATE_WM_DEPTH_STENCIL::DepthTestEnable */
+ if (!pipeline->depth_test_enable)
+ return false;
+
+ /* (((3DSTATE_PS_EXTRA::PixelShaderKillsPixels ||
+ * 3DSTATE_PS_EXTRA::oMask Present to RenderTarget ||
+ * 3DSTATE_PS_BLEND::AlphaToCoverageEnable ||
+ * 3DSTATE_PS_BLEND::AlphaTestEnable ||
+ * 3DSTATE_WM_CHROMAKEY::ChromaKeyKillEnable) &&
+ * 3DSTATE_WM::ForceKillPix != ForceOff &&
+ * ((3DSTATE_WM_DEPTH_STENCIL::DepthWriteEnable &&
+ * 3DSTATE_DEPTH_BUFFER::DEPTH_WRITE_ENABLE) ||
+ * (3DSTATE_WM_DEPTH_STENCIL::Stencil Buffer Write Enable &&
+ * 3DSTATE_DEPTH_BUFFER::STENCIL_WRITE_ENABLE &&
+ * 3DSTATE_STENCIL_BUFFER::STENCIL_BUFFER_ENABLE))) ||
+ * (3DSTATE_PS_EXTRA:: Pixel Shader Computed Depth mode != PSCDEPTH_OFF))
+ */
+ return (pipeline->kill_pixel && (pipeline->writes_depth ||
+ pipeline->writes_stencil)) ||
+ wm_prog_data->computed_depth_mode != PSCDEPTH_OFF;
+}
- if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_DYNAMIC_SCISSOR)
- gen7_cmd_buffer_emit_scissor(cmd_buffer);
+static inline bool
+want_stencil_pma_fix(struct anv_cmd_buffer *cmd_buffer)
+{
+ assert(GEN_GEN == 9);
- genX(cmd_buffer_flush_dynamic_state)(cmd_buffer);
+ /* From the Skylake PRM Vol. 2c CACHE_MODE_1::STC PMA Optimization Enable:
+ *
+ * Clearing this bit will force the STC cache to wait for pending
+ * retirement of pixels at the HZ-read stage and do the STC-test for
+ * Non-promoted, R-computed and Computed depth modes instead of
+ * postponing the STC-test to RCPFE.
+ *
+ * STC_TEST_EN = 3DSTATE_STENCIL_BUFFER::STENCIL_BUFFER_ENABLE &&
+ * 3DSTATE_WM_DEPTH_STENCIL::StencilTestEnable
+ *
+ * STC_WRITE_EN = 3DSTATE_STENCIL_BUFFER::STENCIL_BUFFER_ENABLE &&
+ * (3DSTATE_WM_DEPTH_STENCIL::Stencil Buffer Write Enable &&
+ * 3DSTATE_DEPTH_BUFFER::STENCIL_WRITE_ENABLE)
+ *
+ * COMP_STC_EN = STC_TEST_EN &&
+ * 3DSTATE_PS_EXTRA::PixelShaderComputesStencil
+ *
+ * SW parses the pipeline states to generate the following logical
+ * signal indicating if PMA FIX can be enabled.
+ *
+ * STC_PMA_OPT =
+ * 3DSTATE_WM::ForceThreadDispatch != 1 &&
+ * !(3DSTATE_RASTER::ForceSampleCount != NUMRASTSAMPLES_0) &&
+ * 3DSTATE_DEPTH_BUFFER::SURFACE_TYPE != NULL &&
+ * 3DSTATE_DEPTH_BUFFER::HIZ Enable &&
+ * !(3DSTATE_WM::EDSC_Mode == 2) &&
+ * 3DSTATE_PS_EXTRA::PixelShaderValid &&
+ * !(3DSTATE_WM_HZ_OP::DepthBufferClear ||
+ * 3DSTATE_WM_HZ_OP::DepthBufferResolve ||
+ * 3DSTATE_WM_HZ_OP::Hierarchical Depth Buffer Resolve Enable ||
+ * 3DSTATE_WM_HZ_OP::StencilBufferClear) &&
+ * (COMP_STC_EN || STC_WRITE_EN) &&
+ * ((3DSTATE_PS_EXTRA::PixelShaderKillsPixels ||
+ * 3DSTATE_WM::ForceKillPix == ON ||
+ * 3DSTATE_PS_EXTRA::oMask Present to RenderTarget ||
+ * 3DSTATE_PS_BLEND::AlphaToCoverageEnable ||
+ * 3DSTATE_PS_BLEND::AlphaTestEnable ||
+ * 3DSTATE_WM_CHROMAKEY::ChromaKeyKillEnable) ||
+ * (3DSTATE_PS_EXTRA::Pixel Shader Computed Depth mode != PSCDEPTH_OFF))
+ */
+
+ /* These are always true:
+ * 3DSTATE_WM::ForceThreadDispatch != 1 &&
+ * !(3DSTATE_RASTER::ForceSampleCount != NUMRASTSAMPLES_0)
+ */
+
+ /* We only enable the PMA fix if we know for certain that HiZ is enabled.
+ * If we don't know whether HiZ is enabled or not, we disable the PMA fix
+ * and there is no harm.
+ *
+ * (3DSTATE_DEPTH_BUFFER::SURFACE_TYPE != NULL) &&
+ * 3DSTATE_DEPTH_BUFFER::HIZ Enable
+ */
+ if (!cmd_buffer->state.hiz_enabled)
+ return false;
+
+ /* We can't possibly know if HiZ is enabled without the framebuffer */
+ assert(cmd_buffer->state.framebuffer);
+
+ /* HiZ is enabled so we had better have a depth buffer with HiZ */
+ const struct anv_image_view *ds_iview =
+ anv_cmd_buffer_get_depth_stencil_view(cmd_buffer);
+ assert(ds_iview && ds_iview->image->aux_usage == ISL_AUX_USAGE_HIZ);
+
+ /* 3DSTATE_PS_EXTRA::PixelShaderValid */
+ struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
+ if (!anv_pipeline_has_stage(pipeline, MESA_SHADER_FRAGMENT))
+ return false;
+
+ /* !(3DSTATE_WM::EDSC_Mode == 2) */
+ const struct brw_wm_prog_data *wm_prog_data = get_wm_prog_data(pipeline);
+ if (wm_prog_data->early_fragment_tests)
+ return false;
+
+ /* We never use anv_pipeline for HiZ ops so this is trivially true:
+ * !(3DSTATE_WM_HZ_OP::DepthBufferClear ||
+ * 3DSTATE_WM_HZ_OP::DepthBufferResolve ||
+ * 3DSTATE_WM_HZ_OP::Hierarchical Depth Buffer Resolve Enable ||
+ * 3DSTATE_WM_HZ_OP::StencilBufferClear)
+ */
+
+ /* 3DSTATE_STENCIL_BUFFER::STENCIL_BUFFER_ENABLE &&
+ * 3DSTATE_WM_DEPTH_STENCIL::StencilTestEnable
+ */
+ const bool stc_test_en =
+ (ds_iview->image->aspects & VK_IMAGE_ASPECT_STENCIL_BIT) &&
+ pipeline->stencil_test_enable;
+
+ /* 3DSTATE_STENCIL_BUFFER::STENCIL_BUFFER_ENABLE &&
+ * (3DSTATE_WM_DEPTH_STENCIL::Stencil Buffer Write Enable &&
+ * 3DSTATE_DEPTH_BUFFER::STENCIL_WRITE_ENABLE)
+ */
+ const bool stc_write_en =
+ (ds_iview->image->aspects & VK_IMAGE_ASPECT_STENCIL_BIT) &&
+ pipeline->writes_stencil;
+
+ /* STC_TEST_EN && 3DSTATE_PS_EXTRA::PixelShaderComputesStencil */
+ const bool comp_stc_en = stc_test_en && wm_prog_data->computed_stencil;
+
+ /* COMP_STC_EN || STC_WRITE_EN */
+ if (!(comp_stc_en || stc_write_en))
+ return false;
+
+ /* (3DSTATE_PS_EXTRA::PixelShaderKillsPixels ||
+ * 3DSTATE_WM::ForceKillPix == ON ||
+ * 3DSTATE_PS_EXTRA::oMask Present to RenderTarget ||
+ * 3DSTATE_PS_BLEND::AlphaToCoverageEnable ||
+ * 3DSTATE_PS_BLEND::AlphaTestEnable ||
+ * 3DSTATE_WM_CHROMAKEY::ChromaKeyKillEnable) ||
+ * (3DSTATE_PS_EXTRA::Pixel Shader Computed Depth mode != PSCDEPTH_OFF)
+ */
+ return pipeline->kill_pixel ||
+ wm_prog_data->computed_depth_mode != PSCDEPTH_OFF;
}
void
GENX(COLOR_CALC_STATE_pack)(NULL, cc_state.map, &cc);
if (!cmd_buffer->device->info.has_llc)
- anv_state_clflush(cc_state);
+ anv_state_flush(cc_state);
- anv_batch_emit(&cmd_buffer->batch,
- GENX(3DSTATE_CC_STATE_POINTERS),
- .ColorCalcStatePointer = cc_state.offset,
- .ColorCalcStatePointerValid = true);
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_CC_STATE_POINTERS), ccp) {
+ ccp.ColorCalcStatePointer = cc_state.offset;
+ ccp.ColorCalcStatePointerValid = true;
+ }
}
if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
+ ANV_CMD_DIRTY_RENDER_TARGETS |
ANV_CMD_DIRTY_DYNAMIC_STENCIL_COMPARE_MASK |
ANV_CMD_DIRTY_DYNAMIC_STENCIL_WRITE_MASK)) {
uint32_t wm_depth_stencil_dw[GENX(3DSTATE_WM_DEPTH_STENCIL_length)];
.BackfaceStencilTestMask = d->stencil_compare_mask.back & 0xff,
.BackfaceStencilWriteMask = d->stencil_write_mask.back & 0xff,
+
+ .StencilBufferWriteEnable =
+ (d->stencil_write_mask.front || d->stencil_write_mask.back) &&
+ pipeline->writes_stencil,
};
GENX(3DSTATE_WM_DEPTH_STENCIL_pack)(NULL, wm_depth_stencil_dw,
&wm_depth_stencil);
anv_batch_emit_merge(&cmd_buffer->batch, wm_depth_stencil_dw,
pipeline->gen8.wm_depth_stencil);
+
+ genX(cmd_buffer_enable_pma_fix)(cmd_buffer,
+ want_depth_pma_fix(cmd_buffer));
}
#else
if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_DYNAMIC_BLEND_CONSTANTS) {
GEN9_COLOR_CALC_STATE_pack(NULL, cc_state.map, &cc);
if (!cmd_buffer->device->info.has_llc)
- anv_state_clflush(cc_state);
+ anv_state_flush(cc_state);
- anv_batch_emit(&cmd_buffer->batch,
- GEN9_3DSTATE_CC_STATE_POINTERS,
- .ColorCalcStatePointer = cc_state.offset,
- .ColorCalcStatePointerValid = true);
+ anv_batch_emit(&cmd_buffer->batch, GEN9_3DSTATE_CC_STATE_POINTERS, ccp) {
+ ccp.ColorCalcStatePointer = cc_state.offset;
+ ccp.ColorCalcStatePointerValid = true;
+ }
}
if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
+ ANV_CMD_DIRTY_RENDER_TARGETS |
ANV_CMD_DIRTY_DYNAMIC_STENCIL_COMPARE_MASK |
ANV_CMD_DIRTY_DYNAMIC_STENCIL_WRITE_MASK |
ANV_CMD_DIRTY_DYNAMIC_STENCIL_REFERENCE)) {
.StencilReferenceValue = d->stencil_reference.front & 0xff,
.BackfaceStencilReferenceValue = d->stencil_reference.back & 0xff,
+
+ .StencilBufferWriteEnable =
+ (d->stencil_write_mask.front || d->stencil_write_mask.back) &&
+ pipeline->writes_stencil,
};
GEN9_3DSTATE_WM_DEPTH_STENCIL_pack(NULL, dwords, &wm_depth_stencil);
anv_batch_emit_merge(&cmd_buffer->batch, dwords,
pipeline->gen9.wm_depth_stencil);
+
+ genX(cmd_buffer_enable_pma_fix)(cmd_buffer,
+ want_stencil_pma_fix(cmd_buffer));
}
#endif
if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
ANV_CMD_DIRTY_INDEX_BUFFER)) {
- anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_VF),
- .IndexedDrawCutIndexEnable = pipeline->primitive_restart,
- .CutIndex = cmd_buffer->state.restart_index,
- );
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_VF), vf) {
+ vf.IndexedDrawCutIndexEnable = pipeline->primitive_restart;
+ vf.CutIndex = cmd_buffer->state.restart_index;
+ }
}
cmd_buffer->state.dirty = 0;
cmd_buffer->state.restart_index = restart_index_for_type[indexType];
- anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_INDEX_BUFFER),
- .IndexFormat = vk_to_gen_index_type[indexType],
- .MemoryObjectControlState = GENX(MOCS),
- .BufferStartingAddress = { buffer->bo, buffer->offset + offset },
- .BufferSize = buffer->size - offset);
-
- cmd_buffer->state.dirty |= ANV_CMD_DIRTY_INDEX_BUFFER;
-}
-
-static VkResult
-flush_compute_descriptor_set(struct anv_cmd_buffer *cmd_buffer)
-{
- struct anv_device *device = cmd_buffer->device;
- struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
- struct anv_state surfaces = { 0, }, samplers = { 0, };
- VkResult result;
-
- result = anv_cmd_buffer_emit_samplers(cmd_buffer,
- MESA_SHADER_COMPUTE, &samplers);
- if (result != VK_SUCCESS)
- return result;
- result = anv_cmd_buffer_emit_binding_table(cmd_buffer,
- MESA_SHADER_COMPUTE, &surfaces);
- if (result != VK_SUCCESS)
- return result;
-
- struct anv_state push_state = anv_cmd_buffer_cs_push_constants(cmd_buffer);
-
- const struct brw_cs_prog_data *cs_prog_data = get_cs_prog_data(pipeline);
- const struct brw_stage_prog_data *prog_data = &cs_prog_data->base;
-
- unsigned local_id_dwords = cs_prog_data->local_invocation_id_regs * 8;
- unsigned push_constant_data_size =
- (prog_data->nr_params + local_id_dwords) * 4;
- unsigned reg_aligned_constant_size = ALIGN(push_constant_data_size, 32);
- unsigned push_constant_regs = reg_aligned_constant_size / 32;
-
- if (push_state.alloc_size) {
- anv_batch_emit(&cmd_buffer->batch, GENX(MEDIA_CURBE_LOAD),
- .CURBETotalDataLength = push_state.alloc_size,
- .CURBEDataStartAddress = push_state.offset);
- }
-
- assert(prog_data->total_shared <= 64 * 1024);
- uint32_t slm_size = 0;
- if (prog_data->total_shared > 0) {
- /* slm_size is in 4k increments, but must be a power of 2. */
- slm_size = 4 * 1024;
- while (slm_size < prog_data->total_shared)
- slm_size <<= 1;
- slm_size /= 4 * 1024;
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_INDEX_BUFFER), ib) {
+ ib.IndexFormat = vk_to_gen_index_type[indexType];
+ ib.MemoryObjectControlState = GENX(MOCS);
+ ib.BufferStartingAddress =
+ (struct anv_address) { buffer->bo, buffer->offset + offset };
+ ib.BufferSize = buffer->size - offset;
}
- struct anv_state state =
- anv_state_pool_emit(&device->dynamic_state_pool,
- GENX(INTERFACE_DESCRIPTOR_DATA), 64,
- .KernelStartPointer = pipeline->cs_simd,
- .KernelStartPointerHigh = 0,
- .BindingTablePointer = surfaces.offset,
- .BindingTableEntryCount = 0,
- .SamplerStatePointer = samplers.offset,
- .SamplerCount = 0,
- .ConstantIndirectURBEntryReadLength = push_constant_regs,
- .ConstantURBEntryReadOffset = 0,
- .BarrierEnable = cs_prog_data->uses_barrier,
- .SharedLocalMemorySize = slm_size,
- .NumberofThreadsinGPGPUThreadGroup =
- pipeline->cs_thread_width_max);
-
- uint32_t size = GENX(INTERFACE_DESCRIPTOR_DATA_length) * sizeof(uint32_t);
- anv_batch_emit(&cmd_buffer->batch, GENX(MEDIA_INTERFACE_DESCRIPTOR_LOAD),
- .InterfaceDescriptorTotalLength = size,
- .InterfaceDescriptorDataStartAddress = state.offset);
-
- return VK_SUCCESS;
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_INDEX_BUFFER;
}
-void
-genX(cmd_buffer_flush_compute_state)(struct anv_cmd_buffer *cmd_buffer)
-{
- struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
- const struct brw_cs_prog_data *cs_prog_data = get_cs_prog_data(pipeline);
- VkResult result;
-
- assert(pipeline->active_stages == VK_SHADER_STAGE_COMPUTE_BIT);
-
- bool needs_slm = cs_prog_data->base.total_shared > 0;
- config_l3(cmd_buffer, needs_slm);
-
- if (cmd_buffer->state.current_pipeline != GPGPU) {
-#if GEN_GEN < 10
- /* From the Broadwell PRM, Volume 2a: Instructions, PIPELINE_SELECT:
- *
- * Software must clear the COLOR_CALC_STATE Valid field in
- * 3DSTATE_CC_STATE_POINTERS command prior to send a PIPELINE_SELECT
- * with Pipeline Select set to GPGPU.
- *
- * The internal hardware docs recommend the same workaround for Gen9
- * hardware too.
- */
- anv_batch_emit(&cmd_buffer->batch,
- GENX(3DSTATE_CC_STATE_POINTERS));
-#endif
-
- anv_batch_emit(&cmd_buffer->batch, GENX(PIPELINE_SELECT),
-#if GEN_GEN >= 9
- .MaskBits = 3,
-#endif
- .PipelineSelection = GPGPU);
- cmd_buffer->state.current_pipeline = GPGPU;
- }
-
- if (cmd_buffer->state.compute_dirty & ANV_CMD_DIRTY_PIPELINE)
- anv_batch_emit_batch(&cmd_buffer->batch, &pipeline->batch);
-
- if ((cmd_buffer->state.descriptors_dirty & VK_SHADER_STAGE_COMPUTE_BIT) ||
- (cmd_buffer->state.compute_dirty & ANV_CMD_DIRTY_PIPELINE)) {
- result = flush_compute_descriptor_set(cmd_buffer);
- assert(result == VK_SUCCESS);
- cmd_buffer->state.descriptors_dirty &= ~VK_SHADER_STAGE_COMPUTE_BIT;
- }
-
- cmd_buffer->state.compute_dirty = 0;
-}
+/* Set of stage bits for which are pipelined, i.e. they get queued by the
+ * command streamer for later execution.
+ */
+#define ANV_PIPELINE_STAGE_PIPELINED_BITS \
+ (VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | \
+ VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | \
+ VK_PIPELINE_STAGE_TESSELLATION_CONTROL_SHADER_BIT | \
+ VK_PIPELINE_STAGE_TESSELLATION_EVALUATION_SHADER_BIT | \
+ VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT | \
+ VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | \
+ VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | \
+ VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT | \
+ VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | \
+ VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT | \
+ VK_PIPELINE_STAGE_TRANSFER_BIT | \
+ VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT | \
+ VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT | \
+ VK_PIPELINE_STAGE_ALL_COMMANDS_BIT)
void genX(CmdSetEvent)(
VkCommandBuffer commandBuffer,
ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
ANV_FROM_HANDLE(anv_event, event, _event);
- anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
- .DestinationAddressType = DAT_PPGTT,
- .PostSyncOperation = WriteImmediateData,
- .Address = {
- &cmd_buffer->device->dynamic_state_block_pool.bo,
- event->state.offset
- },
- .ImmediateData = VK_EVENT_SET);
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
+ if (stageMask & ANV_PIPELINE_STAGE_PIPELINED_BITS) {
+ pc.StallAtPixelScoreboard = true;
+ pc.CommandStreamerStallEnable = true;
+ }
+
+ pc.DestinationAddressType = DAT_PPGTT,
+ pc.PostSyncOperation = WriteImmediateData,
+ pc.Address = (struct anv_address) {
+ &cmd_buffer->device->dynamic_state_block_pool.bo,
+ event->state.offset
+ };
+ pc.ImmediateData = VK_EVENT_SET;
+ }
}
void genX(CmdResetEvent)(
ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
ANV_FROM_HANDLE(anv_event, event, _event);
- anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
- .DestinationAddressType = DAT_PPGTT,
- .PostSyncOperation = WriteImmediateData,
- .Address = {
- &cmd_buffer->device->dynamic_state_block_pool.bo,
- event->state.offset
- },
- .ImmediateData = VK_EVENT_RESET);
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
+ if (stageMask & ANV_PIPELINE_STAGE_PIPELINED_BITS) {
+ pc.StallAtPixelScoreboard = true;
+ pc.CommandStreamerStallEnable = true;
+ }
+
+ pc.DestinationAddressType = DAT_PPGTT;
+ pc.PostSyncOperation = WriteImmediateData;
+ pc.Address = (struct anv_address) {
+ &cmd_buffer->device->dynamic_state_block_pool.bo,
+ event->state.offset
+ };
+ pc.ImmediateData = VK_EVENT_RESET;
+ }
}
void genX(CmdWaitEvents)(
for (uint32_t i = 0; i < eventCount; i++) {
ANV_FROM_HANDLE(anv_event, event, pEvents[i]);
- anv_batch_emit(&cmd_buffer->batch, GENX(MI_SEMAPHORE_WAIT),
- .WaitMode = PollingMode,
- .CompareOperation = COMPARE_SAD_EQUAL_SDD,
- .SemaphoreDataDword = VK_EVENT_SET,
- .SemaphoreAddress = {
- &cmd_buffer->device->dynamic_state_block_pool.bo,
- event->state.offset
- });
+ anv_batch_emit(&cmd_buffer->batch, GENX(MI_SEMAPHORE_WAIT), sem) {
+ sem.WaitMode = PollingMode,
+ sem.CompareOperation = COMPARE_SAD_EQUAL_SDD,
+ sem.SemaphoreDataDword = VK_EVENT_SET,
+ sem.SemaphoreAddress = (struct anv_address) {
+ &cmd_buffer->device->dynamic_state_block_pool.bo,
+ event->state.offset
+ };
+ }
}
genX(CmdPipelineBarrier)(commandBuffer, srcStageMask, destStageMask,